More than 2.5 million individuals in the U.S. have severe speech disorders that impair their communication, including the 25% of those who that report that their natural speech cannot be understood. The inability to communicate effectively can have drastic effects on an individual's quality of life, disrupting essential functions such as holding meaningful employment, maintaining relationships with family and friends, or even alerting caregivers to urgent needs. These individuals must rely on some form of augmentative and alternative communication (AAC). Unfortunately, little research has addressed the optimization of the AAC interface through quantitative evaluation. Consequently, users are often limited to slow and inflexible speech output that is insufficient to support effective communication. The long-term goal of the research proposed is to develop and quantitatively evaluate new AAC technology that optimizes speed and flexibility of speech output. If successful, this work could significantly benefit those severely paralyzed individuals experiencing the greatest difficulty in achieving adequate communicative function. Communication rates will be improved by reducing the time it takes users to select a target in two ways. First, the arrangement of phonemic targets will be organized to improve selection rates. Three different approaches will be evaluated: alphabetically, by articulatory features, and through an iterative, quantitative process that reduces the distance between phonemes that commonly occur together. We will test the hypothesis that after training, a quantitatively arranged phonemic interface will provide higher communication rates than phonemic interfaces arranged alphabetically or by articulatory features by evaluating communication rates in healthy users acting as a model of severely paralyzed users. Second, regardless of target arrangement, targets that are likely to be selected next in the sequence will be dynamically enlarged, thus increasing the size of the target and reducing the time it takes to make a selection. We will test the hypothesis that a predictive phonemic interface will provide higher communication rates than a non-predictive phonemic interface during and after training, by evaluating communication rates in both healthy users and severely paralyzed users. These studies will lead to advancements in the development of communication interfaces for AAC devices, eventually providing reliable and fast communication to AAC users, including those severely paralyzed individuals who demonstrate the greatest need for improved communication.